Перегляд за Автор "Bedratiuk, Anna"

Зараз показуємо 1 - 2 з 2
  • Ескіз
    Документ
    Method for obtaining rotation-invariant image representation by removing orientation features from autoencoder latent space
    (Хмельницький національний університет, 2025) Bedratiuk, Anna
    In many computer vision tasks, accurate object recognition is complicated by arbitrary object orientations. Ensuring rotation invariance is critical for improving classification accuracy and reducing errors related to the varying placement of objects. This issue is particularly important in real-world environments, where object orientation is rarely controlled. The goal of this study is to develop a method that allows separating rotational features from the semantic essence of an object, while preserving high classification accuracy after removing orientation-related components. This approach enables the construction of models that remain effective under a wide range of input perspectives, thus improving robustness in practical applications. The proposed method is based on using a convolutional variational autoencoder trained on a dataset of images subjected to various rotation angles. Linear regression is then used to identify those latent components that correlate most strongly with the rotation parameter. These components are removed, and the remaining features are used for classification. Additionally, image reconstruction is performed from the reduced latent vector to visually validate rotation invariance and evaluate the preservation of object shape. Experiments on a synthetically rotated binarized digit dataset (modified MNIST) demonstrated that removing rotationsensitive components led to a classification accuracy decrease of no more than 25–30% across latent space dimensions 3–10 (e.g., normalized accuracy dropped from 1.000 to 0.704 at d = 7). Reconstruction experiments showed that the semantic shape of digits was preserved, while specific orientation information was suppressed. The scientific novelty of this work lies in introducing a simple and reproducible method for removing orientation-related features from the latent space of an autoencoder without modifying the model architecture or introducing specialized regularizers. The practical significance of the method is in reducing the influence of arbitrary object orientation on recognition accuracy, thereby increasing the universality and reliability of vision systems in uncontrolled settings. The proposed approach may be useful for building classifiers capable of handling images with varying or unknown orientations during data collection.
  • Ескіз
    Документ
    Seamless tiling of quasi-periodic textures via an optimal cyclic shift on a discrete torus
    (Хмельницький національний університет, 2026) Bedratiuk, Anna
    In practical computer vision and computer graphics pipelines, it is often necessary to repeatedly replicate a single texture sample to construct a large canvas, background, or regular covering. When the mosaic is not strictly periodic, visible seams appear at the boundaries during repetition, disrupting the perceptual continuity of the texture and often manifesting as a regular grid of artifacts. Such seams not only degrade visual quality but can also alter local gradients and spectral components, which is critical for subsequent processing stages. Common seamless stitching methods increase computational complexity, introduce additional hyperparameters, and modify the local image statistics, which is undesirable in reproducible pipelines and in tasks where the invariance of pixel values is essential. The goal of this work is to propose a simple, reproducible, and computationally efficient method for seam reduction in quasiperiodic textures by selecting an optimal cyclic shift of the pattern that minimizes the energy of mismatch between opposite boundaries. The tile is modeled as a function on the discrete torus ℤ𝑀 × ℤ𝑁. A cyclic shift group 𝐺 = ℤ𝑀 × ℤ𝑁 is introduced, acting as a permutation of pixels. For each shift 𝜏𝑎,𝑏 , the boundary seam energy 𝐸(𝜏𝑎,𝑏 𝐼) is computed in a band of width 𝑤 for opposite boundary pairs, and the minimizing shift is selected. When needed, the evaluation is accelerated via cyclic correlations and FFT. Experiments on synthetic and real textures show that the optimal cyclic shift significantly reduces seam energy and the visual prominence of boundaries during tiling without modifying pixel values. For strictly periodic tiles, the method does not degrade the result. The proposed approach is a lightweight baseline tool for seamless tiling: it does not perform stitching but selects the best cut of the torus. The method is easy to integrate into production pipelines and can be used as a preprocessing step before further processing